With the high cost of memory in, for example, server computer systems, and with the ever-increasing number of processor cores which can be attached to that memory, determining how much memory is enough memory is a significant issue. The typical answer is to look at paging rates between main memory and external storage, and workload response times. If the paging rates between main memory and external storage are high, and response times are unacceptable, then the answer is often to purchase additional memory.
Increasingly though, server computer systems are partitioned with a plurality of logical partitions, such that a single central electronic complex and its main memory may support multiple instances of operating systems. For a given amount (or even for some maximum amount) of memory of the system, it becomes the system administrator's responsibility to determine how much of the physical memory should be assigned to each partition of the partitioned computer system. Similarly, as additional partitions are added, either still-more physical memory needs to be added, or existing partitions need to donate a portion of their memory to the new partitions. Therefore, a need exists for a tool for facilitating determining how much memory each partition requires, and for identifying when the partition requires that memory. Ideally, this determination would be made without unduly impacting performance of the computer system.